SGLang
SGLang is an open-source serving framework hosted under the nonprofit LMSYS organization. It provides high-performance inference for large language and multimodal models, supports low-latency and high-throughput serving, and is intended for research teams and production operators using local or distributed accelerator infrastructure.